I'm sure it's just me, but does anyone out there get real out of co-pilot (the vscode built in one, not the Windows one…good lord, Naming Things™️)?

Beyond useful autocomplete (big `if` blocks, and functions that are similar to other functions) - I've yet to have it fix or write useful blocks…

@rem I have access to Cursor at work. I’ve found it mostly useful for writing the skeletons of test cases (that I then have to fix up), and writing non-production scripts for things (that I often have to fix the guts of, but at least it can find files, distribute work via promise queues, produce some pretty output).

Does it save me time? Who knows. Does it help to motivate me to actually write that spike script I procrastinate on starting? Sure!

@rem the agent mode works alright sometimes. I just used it to track down specifics around how a vue code base was making all its api calls. And then had it verify some api related stuff on the associated rails app. Having it run curl commands and rails console commands to verify stuff itself. This kind of usage I’ve found to help me get oriented in new code bases quicker. I sometimes let it go through and do some coding but usually not entirely thrilled with the output.
@rem I should note I’m having it use Claude 4 model
@rem I recently worked with xState for the first time. Before diving into the documentation, I had a conversation with the chat client in VSCode code, asking it to explain the different concepts of the library. I spent about 30 minutes back-and-forth, asking questions to get a global picture of the library, before diving into the online documentation. This was the first time I did something like this. And it was astonishingly helpful.